Pediatric acute lymphoblastic leukemia (ALL) is the most common pediatric cancer. Although many children are cured by risk stratified therapy, a significant portion either relapse or experience therapy related toxicity. We hypothesize that ALL treatment response is a complex trait which may be partially explained by common genotypic variants. This project will evaluate the association between genotypic variants and therapy outcome on two national randomized clinical trials (CCG-1891 and CCG-1952) of standard risk ALL. This study has four aims. The first aim is to test the impact of polymorphisms, involved in methotrexate (MTX) effect, on treatment outcome in the CCG-1891 sample set. The second aim is to validate associations seen in the CCG-1891 sample set in the CCG-1952 sample set. The third aim is to extend and apply new methods for the analysis of gene-gene interactions to a combined sample set of CCG-1891 and CCG-1952. The fourth aim is to develop a predictive model of ALL relapse risk that includes genotype data. Our prior work has demonstrated in the CCG-1891 sample set that patients homozygous for the MTHFR C677T variant have an increased rate of relapse. We hypothesize that other polymorphisms in the genes mediating MTX effect will modify relapse and toxicity risk. Second, we hypothesize that significant associations seen in CCG-1891 will replicate in CCG-1952. Third, we hypothesize that patterning with recursive partitioning (PRP) will allow identification of polymorphism groups that predict relapse and toxicity. Fourth, we hypothesize that genotype data will improve the clinical utility of predictive models of relapse risk. We propose to test these hypotheses with a nested case control study of 120 relapse patients and 360 patients in continuous remission (CR) on CCG-1891, and of 200 relapse patients and 600 patients in CR on CCG-1952. This application will identify and validate polymorphisms that modify ALL therapy outcome and will rigorously evaluate the additional predictive information captured in genotype data.